1 /*- 2 * BSD LICENSE 3 * 4 * Copyright(c) 2010-2014 Intel Corporation. All rights reserved. 5 * All rights reserved. 6 * 7 * Redistribution and use in source and binary forms, with or without 8 * modification, are permitted provided that the following conditions 9 * are met: 10 * 11 * * Redistributions of source code must retain the above copyright 12 * notice, this list of conditions and the following disclaimer. 13 * * Redistributions in binary form must reproduce the above copyright 14 * notice, this list of conditions and the following disclaimer in 15 * the documentation and/or other materials provided with the 16 * distribution. 17 * * Neither the name of Intel Corporation nor the names of its 18 * contributors may be used to endorse or promote products derived 19 * from this software without specific prior written permission. 20 * 21 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS 22 * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT 23 * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR 24 * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT 25 * OWNER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 26 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT 27 * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, 28 * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY 29 * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 30 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE 31 * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 32 */ 33 34 #include <stdio.h> 35 #include <stdlib.h> 36 #include <stdint.h> 37 #include <inttypes.h> 38 #include <sys/types.h> 39 #include <string.h> 40 #include <sys/queue.h> 41 #include <stdarg.h> 42 #include <errno.h> 43 #include <getopt.h> 44 45 #include <rte_common.h> 46 #include <rte_byteorder.h> 47 #include <rte_log.h> 48 #include <rte_memory.h> 49 #include <rte_memcpy.h> 50 #include <rte_eal.h> 51 #include <rte_launch.h> 52 #include <rte_atomic.h> 53 #include <rte_cycles.h> 54 #include <rte_prefetch.h> 55 #include <rte_lcore.h> 56 #include <rte_per_lcore.h> 57 #include <rte_branch_prediction.h> 58 #include <rte_interrupts.h> 59 #include <rte_random.h> 60 #include <rte_debug.h> 61 #include <rte_ether.h> 62 #include <rte_ethdev.h> 63 #include <rte_mempool.h> 64 #include <rte_mbuf.h> 65 #include <rte_malloc.h> 66 #include <rte_fbk_hash.h> 67 #include <rte_ip.h> 68 69 #define RTE_LOGTYPE_IPv4_MULTICAST RTE_LOGTYPE_USER1 70 71 #define MAX_PORTS 16 72 73 #define MCAST_CLONE_PORTS 2 74 #define MCAST_CLONE_SEGS 2 75 76 #define PKT_MBUF_DATA_SIZE RTE_MBUF_DEFAULT_BUF_SIZE 77 #define NB_PKT_MBUF 8192 78 79 #define HDR_MBUF_DATA_SIZE (2 * RTE_PKTMBUF_HEADROOM) 80 #define NB_HDR_MBUF (NB_PKT_MBUF * MAX_PORTS) 81 82 #define NB_CLONE_MBUF (NB_PKT_MBUF * MCAST_CLONE_PORTS * MCAST_CLONE_SEGS * 2) 83 84 /* allow max jumbo frame 9.5 KB */ 85 #define JUMBO_FRAME_MAX_SIZE 0x2600 86 87 #define MAX_PKT_BURST 32 88 #define BURST_TX_DRAIN_US 100 /* TX drain every ~100us */ 89 90 /* Configure how many packets ahead to prefetch, when reading packets */ 91 #define PREFETCH_OFFSET 3 92 93 /* 94 * Construct Ethernet multicast address from IPv4 multicast address. 95 * Citing RFC 1112, section 6.4: 96 * "An IP host group address is mapped to an Ethernet multicast address 97 * by placing the low-order 23-bits of the IP address into the low-order 98 * 23 bits of the Ethernet multicast address 01-00-5E-00-00-00 (hex)." 99 */ 100 #define ETHER_ADDR_FOR_IPV4_MCAST(x) \ 101 (rte_cpu_to_be_64(0x01005e000000ULL | ((x) & 0x7fffff)) >> 16) 102 103 /* 104 * Configurable number of RX/TX ring descriptors 105 */ 106 #define RTE_TEST_RX_DESC_DEFAULT 128 107 #define RTE_TEST_TX_DESC_DEFAULT 512 108 static uint16_t nb_rxd = RTE_TEST_RX_DESC_DEFAULT; 109 static uint16_t nb_txd = RTE_TEST_TX_DESC_DEFAULT; 110 111 /* ethernet addresses of ports */ 112 static struct ether_addr ports_eth_addr[MAX_PORTS]; 113 114 /* mask of enabled ports */ 115 static uint32_t enabled_port_mask = 0; 116 117 static uint16_t nb_ports; 118 119 static int rx_queue_per_lcore = 1; 120 121 struct mbuf_table { 122 uint16_t len; 123 struct rte_mbuf *m_table[MAX_PKT_BURST]; 124 }; 125 126 #define MAX_RX_QUEUE_PER_LCORE 16 127 #define MAX_TX_QUEUE_PER_PORT 16 128 struct lcore_queue_conf { 129 uint64_t tx_tsc; 130 uint16_t n_rx_queue; 131 uint8_t rx_queue_list[MAX_RX_QUEUE_PER_LCORE]; 132 uint16_t tx_queue_id[MAX_PORTS]; 133 struct mbuf_table tx_mbufs[MAX_PORTS]; 134 } __rte_cache_aligned; 135 static struct lcore_queue_conf lcore_queue_conf[RTE_MAX_LCORE]; 136 137 static struct rte_eth_conf port_conf = { 138 .rxmode = { 139 .max_rx_pkt_len = JUMBO_FRAME_MAX_SIZE, 140 .split_hdr_size = 0, 141 .header_split = 0, /**< Header Split disabled */ 142 .hw_ip_checksum = 0, /**< IP checksum offload disabled */ 143 .hw_vlan_filter = 0, /**< VLAN filtering disabled */ 144 .jumbo_frame = 1, /**< Jumbo Frame Support enabled */ 145 .hw_strip_crc = 1, /**< CRC stripped by hardware */ 146 }, 147 .txmode = { 148 .mq_mode = ETH_MQ_TX_NONE, 149 }, 150 }; 151 152 static struct rte_mempool *packet_pool, *header_pool, *clone_pool; 153 154 155 /* Multicast */ 156 static struct rte_fbk_hash_params mcast_hash_params = { 157 .name = "MCAST_HASH", 158 .entries = 1024, 159 .entries_per_bucket = 4, 160 .socket_id = 0, 161 .hash_func = NULL, 162 .init_val = 0, 163 }; 164 165 struct rte_fbk_hash_table *mcast_hash = NULL; 166 167 struct mcast_group_params { 168 uint32_t ip; 169 uint16_t port_mask; 170 }; 171 172 static struct mcast_group_params mcast_group_table[] = { 173 {IPv4(224,0,0,101), 0x1}, 174 {IPv4(224,0,0,102), 0x2}, 175 {IPv4(224,0,0,103), 0x3}, 176 {IPv4(224,0,0,104), 0x4}, 177 {IPv4(224,0,0,105), 0x5}, 178 {IPv4(224,0,0,106), 0x6}, 179 {IPv4(224,0,0,107), 0x7}, 180 {IPv4(224,0,0,108), 0x8}, 181 {IPv4(224,0,0,109), 0x9}, 182 {IPv4(224,0,0,110), 0xA}, 183 {IPv4(224,0,0,111), 0xB}, 184 {IPv4(224,0,0,112), 0xC}, 185 {IPv4(224,0,0,113), 0xD}, 186 {IPv4(224,0,0,114), 0xE}, 187 {IPv4(224,0,0,115), 0xF}, 188 }; 189 190 #define N_MCAST_GROUPS \ 191 (sizeof (mcast_group_table) / sizeof (mcast_group_table[0])) 192 193 194 /* Send burst of packets on an output interface */ 195 static void 196 send_burst(struct lcore_queue_conf *qconf, uint16_t port) 197 { 198 struct rte_mbuf **m_table; 199 uint16_t n, queueid; 200 int ret; 201 202 queueid = qconf->tx_queue_id[port]; 203 m_table = (struct rte_mbuf **)qconf->tx_mbufs[port].m_table; 204 n = qconf->tx_mbufs[port].len; 205 206 ret = rte_eth_tx_burst(port, queueid, m_table, n); 207 while (unlikely (ret < n)) { 208 rte_pktmbuf_free(m_table[ret]); 209 ret++; 210 } 211 212 qconf->tx_mbufs[port].len = 0; 213 } 214 215 /* Get number of bits set. */ 216 static inline uint32_t 217 bitcnt(uint32_t v) 218 { 219 uint32_t n; 220 221 for (n = 0; v != 0; v &= v - 1, n++) 222 ; 223 224 return n; 225 } 226 227 /** 228 * Create the output multicast packet based on the given input packet. 229 * There are two approaches for creating outgoing packet, though both 230 * are based on data zero-copy idea, they differ in few details: 231 * First one creates a clone of the input packet, e.g - walk though all 232 * segments of the input packet, and for each of them create a new packet 233 * mbuf and attach that new mbuf to the segment (refer to rte_pktmbuf_clone() 234 * for more details). Then new mbuf is allocated for the packet header 235 * and is prepended to the 'clone' mbuf. 236 * Second approach doesn't make a clone, it just increment refcnt for all 237 * input packet segments. Then it allocates new mbuf for the packet header 238 * and prepends it to the input packet. 239 * Basically first approach reuses only input packet's data, but creates 240 * it's own copy of packet's metadata. Second approach reuses both input's 241 * packet data and metadata. 242 * The advantage of first approach - is that each outgoing packet has it's 243 * own copy of metadata, so we can safely modify data pointer of the 244 * input packet. That allows us to skip creation if the output packet for 245 * the last destination port, but instead modify input packet's header inplace, 246 * e.g: for N destination ports we need to invoke mcast_out_pkt (N-1) times. 247 * The advantage of second approach - less work for each outgoing packet, 248 * e.g: we skip "clone" operation completely. Though it comes with a price - 249 * input packet's metadata has to be intact. So for N destination ports we 250 * need to invoke mcast_out_pkt N times. 251 * So for small number of outgoing ports (and segments in the input packet) 252 * first approach will be faster. 253 * As number of outgoing ports (and/or input segments) will grow, 254 * second way will become more preferable. 255 * 256 * @param pkt 257 * Input packet mbuf. 258 * @param use_clone 259 * Control which of the two approaches described above should be used: 260 * - 0 - use second approach: 261 * Don't "clone" input packet. 262 * Prepend new header directly to the input packet 263 * - 1 - use first approach: 264 * Make a "clone" of input packet first. 265 * Prepend new header to the clone of the input packet 266 * @return 267 * - The pointer to the new outgoing packet. 268 * - NULL if operation failed. 269 */ 270 static inline struct rte_mbuf * 271 mcast_out_pkt(struct rte_mbuf *pkt, int use_clone) 272 { 273 struct rte_mbuf *hdr; 274 275 /* Create new mbuf for the header. */ 276 if (unlikely ((hdr = rte_pktmbuf_alloc(header_pool)) == NULL)) 277 return NULL; 278 279 /* If requested, then make a new clone packet. */ 280 if (use_clone != 0 && 281 unlikely ((pkt = rte_pktmbuf_clone(pkt, clone_pool)) == NULL)) { 282 rte_pktmbuf_free(hdr); 283 return NULL; 284 } 285 286 /* prepend new header */ 287 hdr->next = pkt; 288 289 290 /* update header's fields */ 291 hdr->pkt_len = (uint16_t)(hdr->data_len + pkt->pkt_len); 292 hdr->nb_segs = pkt->nb_segs + 1; 293 294 /* copy metadata from source packet*/ 295 hdr->port = pkt->port; 296 hdr->vlan_tci = pkt->vlan_tci; 297 hdr->vlan_tci_outer = pkt->vlan_tci_outer; 298 hdr->tx_offload = pkt->tx_offload; 299 hdr->hash = pkt->hash; 300 301 hdr->ol_flags = pkt->ol_flags; 302 303 __rte_mbuf_sanity_check(hdr, 1); 304 return hdr; 305 } 306 307 /* 308 * Write new Ethernet header to the outgoing packet, 309 * and put it into the outgoing queue for the given port. 310 */ 311 static inline void 312 mcast_send_pkt(struct rte_mbuf *pkt, struct ether_addr *dest_addr, 313 struct lcore_queue_conf *qconf, uint16_t port) 314 { 315 struct ether_hdr *ethdr; 316 uint16_t len; 317 318 /* Construct Ethernet header. */ 319 ethdr = (struct ether_hdr *)rte_pktmbuf_prepend(pkt, (uint16_t)sizeof(*ethdr)); 320 RTE_ASSERT(ethdr != NULL); 321 322 ether_addr_copy(dest_addr, ðdr->d_addr); 323 ether_addr_copy(&ports_eth_addr[port], ðdr->s_addr); 324 ethdr->ether_type = rte_be_to_cpu_16(ETHER_TYPE_IPv4); 325 326 /* Put new packet into the output queue */ 327 len = qconf->tx_mbufs[port].len; 328 qconf->tx_mbufs[port].m_table[len] = pkt; 329 qconf->tx_mbufs[port].len = ++len; 330 331 /* Transmit packets */ 332 if (unlikely(MAX_PKT_BURST == len)) 333 send_burst(qconf, port); 334 } 335 336 /* Multicast forward of the input packet */ 337 static inline void 338 mcast_forward(struct rte_mbuf *m, struct lcore_queue_conf *qconf) 339 { 340 struct rte_mbuf *mc; 341 struct ipv4_hdr *iphdr; 342 uint32_t dest_addr, port_mask, port_num, use_clone; 343 int32_t hash; 344 uint16_t port; 345 union { 346 uint64_t as_int; 347 struct ether_addr as_addr; 348 } dst_eth_addr; 349 350 /* Remove the Ethernet header from the input packet */ 351 iphdr = (struct ipv4_hdr *)rte_pktmbuf_adj(m, (uint16_t)sizeof(struct ether_hdr)); 352 RTE_ASSERT(iphdr != NULL); 353 354 dest_addr = rte_be_to_cpu_32(iphdr->dst_addr); 355 356 /* 357 * Check that it is a valid multicast address and 358 * we have some active ports assigned to it. 359 */ 360 if(!IS_IPV4_MCAST(dest_addr) || 361 (hash = rte_fbk_hash_lookup(mcast_hash, dest_addr)) <= 0 || 362 (port_mask = hash & enabled_port_mask) == 0) { 363 rte_pktmbuf_free(m); 364 return; 365 } 366 367 /* Calculate number of destination ports. */ 368 port_num = bitcnt(port_mask); 369 370 /* Should we use rte_pktmbuf_clone() or not. */ 371 use_clone = (port_num <= MCAST_CLONE_PORTS && 372 m->nb_segs <= MCAST_CLONE_SEGS); 373 374 /* Mark all packet's segments as referenced port_num times */ 375 if (use_clone == 0) 376 rte_pktmbuf_refcnt_update(m, (uint16_t)port_num); 377 378 /* construct destination ethernet address */ 379 dst_eth_addr.as_int = ETHER_ADDR_FOR_IPV4_MCAST(dest_addr); 380 381 for (port = 0; use_clone != port_mask; port_mask >>= 1, port++) { 382 383 /* Prepare output packet and send it out. */ 384 if ((port_mask & 1) != 0) { 385 if (likely ((mc = mcast_out_pkt(m, use_clone)) != NULL)) 386 mcast_send_pkt(mc, &dst_eth_addr.as_addr, 387 qconf, port); 388 else if (use_clone == 0) 389 rte_pktmbuf_free(m); 390 } 391 } 392 393 /* 394 * If we making clone packets, then, for the last destination port, 395 * we can overwrite input packet's metadata. 396 */ 397 if (use_clone != 0) 398 mcast_send_pkt(m, &dst_eth_addr.as_addr, qconf, port); 399 else 400 rte_pktmbuf_free(m); 401 } 402 403 /* Send burst of outgoing packet, if timeout expires. */ 404 static inline void 405 send_timeout_burst(struct lcore_queue_conf *qconf) 406 { 407 uint64_t cur_tsc; 408 uint16_t portid; 409 const uint64_t drain_tsc = (rte_get_tsc_hz() + US_PER_S - 1) / US_PER_S * BURST_TX_DRAIN_US; 410 411 cur_tsc = rte_rdtsc(); 412 if (likely (cur_tsc < qconf->tx_tsc + drain_tsc)) 413 return; 414 415 for (portid = 0; portid < MAX_PORTS; portid++) { 416 if (qconf->tx_mbufs[portid].len != 0) 417 send_burst(qconf, portid); 418 } 419 qconf->tx_tsc = cur_tsc; 420 } 421 422 /* main processing loop */ 423 static int 424 main_loop(__rte_unused void *dummy) 425 { 426 struct rte_mbuf *pkts_burst[MAX_PKT_BURST]; 427 unsigned lcore_id; 428 int i, j, nb_rx; 429 uint16_t portid; 430 struct lcore_queue_conf *qconf; 431 432 lcore_id = rte_lcore_id(); 433 qconf = &lcore_queue_conf[lcore_id]; 434 435 436 if (qconf->n_rx_queue == 0) { 437 RTE_LOG(INFO, IPv4_MULTICAST, "lcore %u has nothing to do\n", 438 lcore_id); 439 return 0; 440 } 441 442 RTE_LOG(INFO, IPv4_MULTICAST, "entering main loop on lcore %u\n", 443 lcore_id); 444 445 for (i = 0; i < qconf->n_rx_queue; i++) { 446 447 portid = qconf->rx_queue_list[i]; 448 RTE_LOG(INFO, IPv4_MULTICAST, " -- lcoreid=%u portid=%d\n", 449 lcore_id, portid); 450 } 451 452 while (1) { 453 454 /* 455 * Read packet from RX queues 456 */ 457 for (i = 0; i < qconf->n_rx_queue; i++) { 458 459 portid = qconf->rx_queue_list[i]; 460 nb_rx = rte_eth_rx_burst(portid, 0, pkts_burst, 461 MAX_PKT_BURST); 462 463 /* Prefetch first packets */ 464 for (j = 0; j < PREFETCH_OFFSET && j < nb_rx; j++) { 465 rte_prefetch0(rte_pktmbuf_mtod( 466 pkts_burst[j], void *)); 467 } 468 469 /* Prefetch and forward already prefetched packets */ 470 for (j = 0; j < (nb_rx - PREFETCH_OFFSET); j++) { 471 rte_prefetch0(rte_pktmbuf_mtod(pkts_burst[ 472 j + PREFETCH_OFFSET], void *)); 473 mcast_forward(pkts_burst[j], qconf); 474 } 475 476 /* Forward remaining prefetched packets */ 477 for (; j < nb_rx; j++) { 478 mcast_forward(pkts_burst[j], qconf); 479 } 480 } 481 482 /* Send out packets from TX queues */ 483 send_timeout_burst(qconf); 484 } 485 } 486 487 /* display usage */ 488 static void 489 print_usage(const char *prgname) 490 { 491 printf("%s [EAL options] -- -p PORTMASK [-q NQ]\n" 492 " -p PORTMASK: hexadecimal bitmask of ports to configure\n" 493 " -q NQ: number of queue (=ports) per lcore (default is 1)\n", 494 prgname); 495 } 496 497 static uint32_t 498 parse_portmask(const char *portmask) 499 { 500 char *end = NULL; 501 unsigned long pm; 502 503 /* parse hexadecimal string */ 504 pm = strtoul(portmask, &end, 16); 505 if ((portmask[0] == '\0') || (end == NULL) || (*end != '\0')) 506 return 0; 507 508 return (uint32_t)pm; 509 } 510 511 static int 512 parse_nqueue(const char *q_arg) 513 { 514 char *end = NULL; 515 unsigned long n; 516 517 /* parse numerical string */ 518 errno = 0; 519 n = strtoul(q_arg, &end, 0); 520 if (errno != 0 || end == NULL || *end != '\0' || 521 n == 0 || n >= MAX_RX_QUEUE_PER_LCORE) 522 return -1; 523 524 return n; 525 } 526 527 /* Parse the argument given in the command line of the application */ 528 static int 529 parse_args(int argc, char **argv) 530 { 531 int opt, ret; 532 char **argvopt; 533 int option_index; 534 char *prgname = argv[0]; 535 static struct option lgopts[] = { 536 {NULL, 0, 0, 0} 537 }; 538 539 argvopt = argv; 540 541 while ((opt = getopt_long(argc, argvopt, "p:q:", 542 lgopts, &option_index)) != EOF) { 543 544 switch (opt) { 545 /* portmask */ 546 case 'p': 547 enabled_port_mask = parse_portmask(optarg); 548 if (enabled_port_mask == 0) { 549 printf("invalid portmask\n"); 550 print_usage(prgname); 551 return -1; 552 } 553 break; 554 555 /* nqueue */ 556 case 'q': 557 rx_queue_per_lcore = parse_nqueue(optarg); 558 if (rx_queue_per_lcore < 0) { 559 printf("invalid queue number\n"); 560 print_usage(prgname); 561 return -1; 562 } 563 break; 564 565 default: 566 print_usage(prgname); 567 return -1; 568 } 569 } 570 571 if (optind >= 0) 572 argv[optind-1] = prgname; 573 574 ret = optind-1; 575 optind = 1; /* reset getopt lib */ 576 return ret; 577 } 578 579 static void 580 print_ethaddr(const char *name, struct ether_addr *eth_addr) 581 { 582 char buf[ETHER_ADDR_FMT_SIZE]; 583 ether_format_addr(buf, ETHER_ADDR_FMT_SIZE, eth_addr); 584 printf("%s%s", name, buf); 585 } 586 587 static int 588 init_mcast_hash(void) 589 { 590 uint32_t i; 591 592 mcast_hash_params.socket_id = rte_socket_id(); 593 mcast_hash = rte_fbk_hash_create(&mcast_hash_params); 594 if (mcast_hash == NULL){ 595 return -1; 596 } 597 598 for (i = 0; i < N_MCAST_GROUPS; i ++){ 599 if (rte_fbk_hash_add_key(mcast_hash, 600 mcast_group_table[i].ip, 601 mcast_group_table[i].port_mask) < 0) { 602 return -1; 603 } 604 } 605 606 return 0; 607 } 608 609 /* Check the link status of all ports in up to 9s, and print them finally */ 610 static void 611 check_all_ports_link_status(uint16_t port_num, uint32_t port_mask) 612 { 613 #define CHECK_INTERVAL 100 /* 100ms */ 614 #define MAX_CHECK_TIME 90 /* 9s (90 * 100ms) in total */ 615 uint16_t portid; 616 uint8_t count, all_ports_up, print_flag = 0; 617 struct rte_eth_link link; 618 619 printf("\nChecking link status"); 620 fflush(stdout); 621 for (count = 0; count <= MAX_CHECK_TIME; count++) { 622 all_ports_up = 1; 623 for (portid = 0; portid < port_num; portid++) { 624 if ((port_mask & (1 << portid)) == 0) 625 continue; 626 memset(&link, 0, sizeof(link)); 627 rte_eth_link_get_nowait(portid, &link); 628 /* print link status if flag set */ 629 if (print_flag == 1) { 630 if (link.link_status) 631 printf( 632 "Port%d Link Up. Speed %u Mbps - %s\n", 633 portid, link.link_speed, 634 (link.link_duplex == ETH_LINK_FULL_DUPLEX) ? 635 ("full-duplex") : ("half-duplex\n")); 636 else 637 printf("Port %d Link Down\n", portid); 638 continue; 639 } 640 /* clear all_ports_up flag if any link down */ 641 if (link.link_status == ETH_LINK_DOWN) { 642 all_ports_up = 0; 643 break; 644 } 645 } 646 /* after finally printing all link status, get out */ 647 if (print_flag == 1) 648 break; 649 650 if (all_ports_up == 0) { 651 printf("."); 652 fflush(stdout); 653 rte_delay_ms(CHECK_INTERVAL); 654 } 655 656 /* set the print_flag if all ports up or timeout */ 657 if (all_ports_up == 1 || count == (MAX_CHECK_TIME - 1)) { 658 print_flag = 1; 659 printf("done\n"); 660 } 661 } 662 } 663 664 int 665 main(int argc, char **argv) 666 { 667 struct lcore_queue_conf *qconf; 668 struct rte_eth_dev_info dev_info; 669 struct rte_eth_txconf *txconf; 670 int ret; 671 uint16_t queueid; 672 unsigned lcore_id = 0, rx_lcore_id = 0; 673 uint32_t n_tx_queue, nb_lcores; 674 uint16_t portid; 675 676 /* init EAL */ 677 ret = rte_eal_init(argc, argv); 678 if (ret < 0) 679 rte_exit(EXIT_FAILURE, "Invalid EAL parameters\n"); 680 argc -= ret; 681 argv += ret; 682 683 /* parse application arguments (after the EAL ones) */ 684 ret = parse_args(argc, argv); 685 if (ret < 0) 686 rte_exit(EXIT_FAILURE, "Invalid IPV4_MULTICAST parameters\n"); 687 688 /* create the mbuf pools */ 689 packet_pool = rte_pktmbuf_pool_create("packet_pool", NB_PKT_MBUF, 32, 690 0, PKT_MBUF_DATA_SIZE, rte_socket_id()); 691 692 if (packet_pool == NULL) 693 rte_exit(EXIT_FAILURE, "Cannot init packet mbuf pool\n"); 694 695 header_pool = rte_pktmbuf_pool_create("header_pool", NB_HDR_MBUF, 32, 696 0, HDR_MBUF_DATA_SIZE, rte_socket_id()); 697 698 if (header_pool == NULL) 699 rte_exit(EXIT_FAILURE, "Cannot init header mbuf pool\n"); 700 701 clone_pool = rte_pktmbuf_pool_create("clone_pool", NB_CLONE_MBUF, 32, 702 0, 0, rte_socket_id()); 703 704 if (clone_pool == NULL) 705 rte_exit(EXIT_FAILURE, "Cannot init clone mbuf pool\n"); 706 707 nb_ports = rte_eth_dev_count(); 708 if (nb_ports == 0) 709 rte_exit(EXIT_FAILURE, "No physical ports!\n"); 710 if (nb_ports > MAX_PORTS) 711 nb_ports = MAX_PORTS; 712 713 nb_lcores = rte_lcore_count(); 714 715 /* initialize all ports */ 716 for (portid = 0; portid < nb_ports; portid++) { 717 /* skip ports that are not enabled */ 718 if ((enabled_port_mask & (1 << portid)) == 0) { 719 printf("Skipping disabled port %d\n", portid); 720 continue; 721 } 722 723 qconf = &lcore_queue_conf[rx_lcore_id]; 724 725 /* limit the frame size to the maximum supported by NIC */ 726 rte_eth_dev_info_get(portid, &dev_info); 727 port_conf.rxmode.max_rx_pkt_len = RTE_MIN( 728 dev_info.max_rx_pktlen, port_conf.rxmode.max_rx_pkt_len); 729 730 /* get the lcore_id for this port */ 731 while (rte_lcore_is_enabled(rx_lcore_id) == 0 || 732 qconf->n_rx_queue == (unsigned)rx_queue_per_lcore) { 733 734 rx_lcore_id ++; 735 qconf = &lcore_queue_conf[rx_lcore_id]; 736 737 if (rx_lcore_id >= RTE_MAX_LCORE) 738 rte_exit(EXIT_FAILURE, "Not enough cores\n"); 739 } 740 qconf->rx_queue_list[qconf->n_rx_queue] = portid; 741 qconf->n_rx_queue++; 742 743 /* init port */ 744 printf("Initializing port %d on lcore %u... ", portid, 745 rx_lcore_id); 746 fflush(stdout); 747 748 n_tx_queue = nb_lcores; 749 if (n_tx_queue > MAX_TX_QUEUE_PER_PORT) 750 n_tx_queue = MAX_TX_QUEUE_PER_PORT; 751 ret = rte_eth_dev_configure(portid, 1, (uint16_t)n_tx_queue, 752 &port_conf); 753 if (ret < 0) 754 rte_exit(EXIT_FAILURE, "Cannot configure device: err=%d, port=%d\n", 755 ret, portid); 756 757 ret = rte_eth_dev_adjust_nb_rx_tx_desc(portid, &nb_rxd, 758 &nb_txd); 759 if (ret < 0) 760 rte_exit(EXIT_FAILURE, 761 "Cannot adjust number of descriptors: err=%d, port=%d\n", 762 ret, portid); 763 764 rte_eth_macaddr_get(portid, &ports_eth_addr[portid]); 765 print_ethaddr(" Address:", &ports_eth_addr[portid]); 766 printf(", "); 767 768 /* init one RX queue */ 769 queueid = 0; 770 printf("rxq=%hu ", queueid); 771 fflush(stdout); 772 ret = rte_eth_rx_queue_setup(portid, queueid, nb_rxd, 773 rte_eth_dev_socket_id(portid), 774 NULL, 775 packet_pool); 776 if (ret < 0) 777 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, port=%d\n", 778 ret, portid); 779 780 /* init one TX queue per couple (lcore,port) */ 781 queueid = 0; 782 783 RTE_LCORE_FOREACH(lcore_id) { 784 if (rte_lcore_is_enabled(lcore_id) == 0) 785 continue; 786 printf("txq=%u,%hu ", lcore_id, queueid); 787 fflush(stdout); 788 789 txconf = &dev_info.default_txconf; 790 txconf->txq_flags = 0; 791 ret = rte_eth_tx_queue_setup(portid, queueid, nb_txd, 792 rte_lcore_to_socket_id(lcore_id), txconf); 793 if (ret < 0) 794 rte_exit(EXIT_FAILURE, "rte_eth_tx_queue_setup: err=%d, " 795 "port=%d\n", ret, portid); 796 797 qconf = &lcore_queue_conf[lcore_id]; 798 qconf->tx_queue_id[portid] = queueid; 799 queueid++; 800 } 801 802 /* Start device */ 803 ret = rte_eth_dev_start(portid); 804 if (ret < 0) 805 rte_exit(EXIT_FAILURE, "rte_eth_dev_start: err=%d, port=%d\n", 806 ret, portid); 807 808 printf("done:\n"); 809 } 810 811 check_all_ports_link_status(nb_ports, enabled_port_mask); 812 813 /* initialize the multicast hash */ 814 int retval = init_mcast_hash(); 815 if (retval != 0) 816 rte_exit(EXIT_FAILURE, "Cannot build the multicast hash\n"); 817 818 /* launch per-lcore init on every lcore */ 819 rte_eal_mp_remote_launch(main_loop, NULL, CALL_MASTER); 820 RTE_LCORE_FOREACH_SLAVE(lcore_id) { 821 if (rte_eal_wait_lcore(lcore_id) < 0) 822 return -1; 823 } 824 825 return 0; 826 } 827